Greedy Layer-Wise Neural Net.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Regression Problem with Greedy Layer-Wise Trained Neural Network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Libraries and settings"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**1.1 Importing libraries**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import numpy as np\n",
    "import os\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import mean_absolute_error\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from keras.layers import Dense, Dropout, Flatten\n",
    "from keras.models import Sequential\n",
    "from keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "os.chdir('C:/Users/Nicolas/Documents/Data/electric-motor-'\\\n",
    "         'temperature')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**1.2 Plot theme**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "theme_color = '#8FCB9B'\n",
    "\n",
    "def set_theme():\n",
    "    return ax.set_facecolor(theme_color),\\\n",
    "    ax.grid(True, alpha=4e-1, color='white'),\\\n",
    "    ax.set_axisbelow(True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Importing the data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**2.1 Importing**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('pmsm_temperature_data.csv').astype(np.float32)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**2.2 Exploring**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "2.2.1 Head"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ambient</th>\n",
       "      <th>coolant</th>\n",
       "      <th>u_d</th>\n",
       "      <th>u_q</th>\n",
       "      <th>motor_speed</th>\n",
       "      <th>torque</th>\n",
       "      <th>i_d</th>\n",
       "      <th>i_q</th>\n",
       "      <th>pm</th>\n",
       "      <th>stator_yoke</th>\n",
       "      <th>stator_tooth</th>\n",
       "      <th>stator_winding</th>\n",
       "      <th>profile_id</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.752143</td>\n",
       "      <td>-1.118446</td>\n",
       "      <td>0.327935</td>\n",
       "      <td>-1.297858</td>\n",
       "      <td>-1.222428</td>\n",
       "      <td>-0.250182</td>\n",
       "      <td>1.029572</td>\n",
       "      <td>-0.245860</td>\n",
       "      <td>-2.522071</td>\n",
       "      <td>-1.831422</td>\n",
       "      <td>-2.066143</td>\n",
       "      <td>-2.018033</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-0.771263</td>\n",
       "      <td>-1.117021</td>\n",
       "      <td>0.329665</td>\n",
       "      <td>-1.297686</td>\n",
       "      <td>-1.222429</td>\n",
       "      <td>-0.249133</td>\n",
       "      <td>1.029509</td>\n",
       "      <td>-0.245832</td>\n",
       "      <td>-2.522418</td>\n",
       "      <td>-1.830969</td>\n",
       "      <td>-2.064859</td>\n",
       "      <td>-2.017631</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-0.782892</td>\n",
       "      <td>-1.116681</td>\n",
       "      <td>0.332772</td>\n",
       "      <td>-1.301822</td>\n",
       "      <td>-1.222428</td>\n",
       "      <td>-0.249431</td>\n",
       "      <td>1.029448</td>\n",
       "      <td>-0.245818</td>\n",
       "      <td>-2.522673</td>\n",
       "      <td>-1.830400</td>\n",
       "      <td>-2.064073</td>\n",
       "      <td>-2.017344</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-0.780935</td>\n",
       "      <td>-1.116764</td>\n",
       "      <td>0.333700</td>\n",
       "      <td>-1.301852</td>\n",
       "      <td>-1.222430</td>\n",
       "      <td>-0.248636</td>\n",
       "      <td>1.032845</td>\n",
       "      <td>-0.246955</td>\n",
       "      <td>-2.521639</td>\n",
       "      <td>-1.830333</td>\n",
       "      <td>-2.063137</td>\n",
       "      <td>-2.017632</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-0.774043</td>\n",
       "      <td>-1.116775</td>\n",
       "      <td>0.335206</td>\n",
       "      <td>-1.303118</td>\n",
       "      <td>-1.222429</td>\n",
       "      <td>-0.248701</td>\n",
       "      <td>1.031807</td>\n",
       "      <td>-0.246610</td>\n",
       "      <td>-2.521900</td>\n",
       "      <td>-1.830498</td>\n",
       "      <td>-2.062795</td>\n",
       "      <td>-2.018145</td>\n",
       "      <td>4.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    ambient   coolant       u_d       u_q  motor_speed    torque       i_d  \\\n",
       "0 -0.752143 -1.118446  0.327935 -1.297858    -1.222428 -0.250182  1.029572   \n",
       "1 -0.771263 -1.117021  0.329665 -1.297686    -1.222429 -0.249133  1.029509   \n",
       "2 -0.782892 -1.116681  0.332772 -1.301822    -1.222428 -0.249431  1.029448   \n",
       "3 -0.780935 -1.116764  0.333700 -1.301852    -1.222430 -0.248636  1.032845   \n",
       "4 -0.774043 -1.116775  0.335206 -1.303118    -1.222429 -0.248701  1.031807   \n",
       "\n",
       "        i_q        pm  stator_yoke  stator_tooth  stator_winding  profile_id  \n",
       "0 -0.245860 -2.522071    -1.831422     -2.066143       -2.018033         4.0  \n",
       "1 -0.245832 -2.522418    -1.830969     -2.064859       -2.017631         4.0  \n",
       "2 -0.245818 -2.522673    -1.830400     -2.064073       -2.017344         4.0  \n",
       "3 -0.246955 -2.521639    -1.830333     -2.063137       -2.017632         4.0  \n",
       "4 -0.246610 -2.521900    -1.830498     -2.062795       -2.018145         4.0  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "2.2.2 Profile ID"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "20.0    43970\n",
       "6.0     40387\n",
       "65.0    40093\n",
       "66.0    36475\n",
       "27.0    35360\n",
       "Name: profile_id, dtype: int64"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['profile_id'].value_counts().head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.loc[df['profile_id'] == 6].drop('profile_id', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "We now have 40,387 rows and 12 columns.\n"
     ]
    }
   ],
   "source": [
    "print(f'We now have {df.shape[0]:,} rows and {df.shape[1]} columns.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Visualizing distributions of the target variables"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**3.1 Stator tooth**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(8, 5))\n",
    "set_theme()\n",
    "sns.distplot(df.iloc[:, -2], color='#456990', \n",
    "             bins=30, kde=False)\n",
    "ax.set_yticklabels(['{:,}'.format(x) for x in np.array(ax.get_yticks(), \n",
    "                                                       dtype=int)])\n",
    "plt.title(\"Distribution of values in column '{}'\".format(df.columns[-3]))\n",
    "plt.xlabel(None)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Processing data for predictive modeling"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**4.1 Separating the predictor and target variables**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df.drop(['stator_yoke', 'stator_tooth', 'stator_winding'], axis=1).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = df['stator_tooth'].values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 4.2 Rescaling the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "scaler = MinMaxScaler()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = scaler.fit_transform(X)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**4.2 Train/test split**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=2e-1)             "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5. Greedy Layer-Wise Neural Net vs Dense Neural Net"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.1 Greedy Layer-Wise Neural Net"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We're going to start off with a single-layer neural net. Then we will add identical layers, trained 'greedily', up to 10."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.1.1 Building the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential([\n",
    "    Dense(48, input_shape=(X.shape[1],), trainable=True, name='first'),\n",
    "    Dense(1, activation='linear')\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.1.2 Compiling the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(optimizer='adam', loss='mae')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.1.3 Training the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 32309 samples, validate on 8078 samples\n",
      "Epoch 1/10\n",
      "32309/32309 [==============================] - 17s 542us/step - loss: 0.1818 - val_loss: 0.1655\n",
      "Epoch 2/10\n",
      "32309/32309 [==============================] - 13s 400us/step - loss: 0.1596 - val_loss: 0.1559\n",
      "Epoch 3/10\n",
      "32309/32309 [==============================] - 13s 394us/step - loss: 0.1587 - val_loss: 0.1566\n",
      "Epoch 4/10\n",
      "32309/32309 [==============================] - 12s 383us/step - loss: 0.1574 - val_loss: 0.1571\n",
      "Epoch 5/10\n",
      "32309/32309 [==============================] - 13s 407us/step - loss: 0.1575 - val_loss: 0.1552\n",
      "Epoch 6/10\n",
      "32309/32309 [==============================] - 12s 374us/step - loss: 0.1569 - val_loss: 0.1626\n",
      "Epoch 7/10\n",
      "32309/32309 [==============================] - 12s 367us/step - loss: 0.1565 - val_loss: 0.1550\n",
      "Epoch 8/10\n",
      "32309/32309 [==============================] - 12s 384us/step - loss: 0.1565 - val_loss: 0.1561\n",
      "Epoch 9/10\n",
      "32309/32309 [==============================] - 12s 379us/step - loss: 0.1559 - val_loss: 0.1554\n",
      "Epoch 10/10\n",
      "32309/32309 [==============================] - 12s 380us/step - loss: 0.1559 - val_loss: 0.1553\n"
     ]
    }
   ],
   "source": [
    "first = model.fit(X_train, y_train,\n",
    "                  epochs=10,\n",
    "                  validation_data=[X_test, y_test],\n",
    "                  batch_size=8)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.1.4 Making predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = model.predict(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.1.5 Evaluating predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.15531653"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mae_first = mean_absolute_error(y_test, y_pred)\n",
    "mae_first"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.1.6 Making a metrics dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>MAE Train</th>\n",
       "      <th>MAE Test</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>One Layer</th>\n",
       "      <td>0.155934</td>\n",
       "      <td>0.155317</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           MAE Train  MAE Test\n",
       "One Layer   0.155934  0.155317"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metrics = pd.DataFrame([first.history['loss'][-1], mae_first], \n",
    "                       columns=['One Layer'], index=['MAE Train', 'MAE Test']).T\n",
    "metrics"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.1.7 Plotting the metrics"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.2 Adding More Layers"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.2.1 Function to evaluate the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluate_model(model, X_train, X_test, y_train, y_test):\n",
    "    train_loss = model.evaluate(X_train, y_train, verbose=0)\n",
    "    test_loss = model.evaluate(X_test, y_test, verbose=0)\n",
    "    return train_loss, test_loss"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.2.2 Function to add layers and evaluate the model's performance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "def add_layer():\n",
    "    output_layer = model.layers[-1]\n",
    "    model.pop()\n",
    "    for layer in model.layers:\n",
    "        layer.trainable = False\n",
    "    model.add(Dense(48, activation='relu', kernel_initializer='he_uniform'))\n",
    "    model.add(output_layer)\n",
    "    \n",
    "    model.fit(X_train, y_train,\n",
    "              epochs=10,\n",
    "              batch_size=8,\n",
    "              verbose=0)\n",
    "    \n",
    "    train_mae, test_mae = evaluate_model(model, X_train, X_test, y_train, y_test)\n",
    "    return train_mae, test_mae"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.2.3 Training the models, up to 10 layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss: 0.040023933932715104 — Test loss: 0.04025506637861349\n",
      "Train loss: 0.0441103787772366 — Test loss: 0.044619352850296556\n",
      "Train loss: 0.0318532315782702 — Test loss: 0.03226936412650071\n",
      "Train loss: 0.031455077156906486 — Test loss: 0.03190078808344748\n",
      "Train loss: 0.02738665429300878 — Test loss: 0.027713152854817506\n",
      "Train loss: 0.019409112999050092 — Test loss: 0.0197878060336694\n",
      "Train loss: 0.018066894518941747 — Test loss: 0.018076957351020596\n",
      "Train loss: 0.01656917078584998 — Test loss: 0.016372689290844247\n",
      "Train loss: 0.016209089571747937 — Test loss: 0.015990308411677645\n",
      "Train loss: 0.021157031696421337 — Test loss: 0.02094251515127499\n"
     ]
    }
   ],
   "source": [
    "train_loss_history, test_loss_history = [], []\n",
    "for i in range(10):\n",
    "    train_loss, test_loss = add_layer()\n",
    "    train_loss_history.append(train_loss)\n",
    "    test_loss_history.append(test_loss)\n",
    "    print('Train loss: {} — Test loss: {}'.format(train_loss, test_loss))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.2.4 Plotting the metrics history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(8, 5))\n",
    "set_theme()\n",
    "\n",
    "ax.plot(np.arange(1, 11), train_loss_history, color='orange', marker='o', \n",
    "       markeredgecolor='k', alpha=1, markersize=10)\n",
    "\n",
    "ax.plot(np.arange(1, 11), test_loss_history, color='green', marker='o', \n",
    "       markeredgecolor='k', alpha=1, markersize=10)\n",
    "\n",
    "plt.legend(['Train Loss', 'Test Loss'], framealpha=.6)\n",
    "plt.title('MAE Depending on Number of Layers')\n",
    "plt.xlabel('Number of Layers')\n",
    "plt.ylabel('MAE')\n",
    "plt.xticks(np.arange(1, 11))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that the best model is the one with 9 layers. So we will compare it with a fully-connected neural network trained normally. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.3 Comparing with a Fully-Connected Neural Net (9 layers)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.3.1 Building the models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential([\n",
    "    Dense(48, input_shape=(X.shape[1],)),\n",
    "    Dense(48, activation='relu', kernel_initializer='he_uniform'),\n",
    "    Dense(48, activation='relu', kernel_initializer='he_uniform'),\n",
    "    Dense(48, activation='relu', kernel_initializer='he_uniform'),\n",
    "    Dense(48, activation='relu', kernel_initializer='he_uniform'),\n",
    "    Dense(48, activation='relu', kernel_initializer='he_uniform'),\n",
    "    Dense(48, activation='relu', kernel_initializer='he_uniform'),\n",
    "    Dense(48, activation='relu', kernel_initializer='he_uniform'),\n",
    "    Dense(48, activation='relu', kernel_initializer='he_uniform'),\n",
    "    Dense(48, activation='relu', kernel_initializer='he_uniform'),\n",
    "    Dense(1, activation='linear')\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.3.2 Compiling the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(optimizer='adam', loss='mae')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.3.3 Fitting the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "_ = model.fit(X_train, y_train,\n",
    "              epochs=9*10,\n",
    "              batch_size=8,\n",
    "              verbose=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.3.4 Evaluating the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_mae_fc, test_mae_fc = evaluate_model(model, X_train, X_test, y_train, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "5.3.5 Plotting the performance of all models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(8, 5))\n",
    "set_theme()\n",
    "\n",
    "ax.plot(np.arange(1, 11), train_loss_history, color='orange', marker='o', \n",
    "       mec='k', alpha=1, ms=10, ls='--', visible=True)\n",
    "\n",
    "ax.plot(np.arange(1, 11), test_loss_history, color='green', marker='o', \n",
    "       mec='k', alpha=1, ms=10, ls='--', visible=True)\n",
    "\n",
    "ax.plot([1, 10], [train_mae_fc, train_mae_fc], color='red', marker='s', \n",
    "       mec='k', alpha=1, ms=10, ls='-', visible=True)\n",
    "\n",
    "ax.plot([1, 10], [test_mae_fc, test_mae_fc], color='yellow', marker='s', \n",
    "       mec='k', alpha=1, ms=10, ls='-', visible=True)\n",
    "\n",
    "plt.legend(['Train Loss', 'Test Loss', '6 Layer Train Loss', '6 Layer Test Loss'], framealpha=.6)\n",
    "plt.title('MAE Depending on Number of Layers')\n",
    "plt.xlabel('Number of Layers')\n",
    "plt.ylabel('MAE')\n",
    "plt.xticks(np.arange(1, 11))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The greedy layer-wise neural net strongly outperforms the backprop trained neural net. "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}